This is an implementation of the paper "Journey to the center of the words: Word weighting scheme based on the geometry of word embeddings."
Setup:
- Install the python packages given in requirements.txt.
- Keep the STS datasets in datasets folder.
- Keep the word embedding files (GloVe, Word2Vec) in data folder. Instructions are given in data folder on how to download the word embedding files.
Run:
- Simply, run the localnorm.py file. Current code works with Glove embeddigs.
- For Word2vec, uncomment lines #201 - #209 and comment out lines #198 - #199.